Ezrin Promotes Morphogenesis of Apical Microvilli and Basal Infoldings in Retinal Pigment Epithelium

Reagents were from Sigma Chemical Co. or GIBCO BRL, unless otherwise stated. mAb3056 to the COOH-terminal domain of ezrin was purchased from Chemicon. Polyclonal ezrin antibody was a generous gift from Dr. Monique Arpin (Curie Institute, Paris, France). Antiezrin polyclonal antibody (pAb) C-19 directed to the peptide 566–584 of human ezrin (conserved in all ERM proteins) was purchased from Santa Cruz Biotechnology, Inc. Affinity-purified polyclonal moesin antibody was a generous gift from Dr. Anthony Bretscher (Cornell University, Ithaca, NY). Polyclonal antibodies to radixin (457 and 220) were generously provided by Dr. Frank Solomon (MIT, Cambridge, MA). mAb to vesicular stomatitis virus glycoprotein G (VSVG), clone P5D4, raised against the 11-amino acid COOH terminus of the VSVG was previously described (Kreis 1986). Secondary antibodies were purchased from Cappel Laboratories or Jackson ImmunoResearch Laboratories, Inc.

RPE-J cells were originally obtained from rat RPE by immortalization with temperature sensitive SV40 T antigen (Nabi et al. 1993). The cells were cultured at the permissive temperature of 32°C in DME medium supplemented with 4% heat-inactivated CELLect Gold FCS (ICN Pharmaceuticals, Inc.), l-glutamine, nonessential amino acids, and antibiotics. To promote differentiation, cells were plated on polycarbonate Transwell filters (Costar) coated with a thin layer of Matrigel (Collaborative Research). The cells were then cultured in growth medium supplemented with 10⁻⁸ M retinoic acid for 6 d, and then switched to 40°C, the nonpermissive temperature for transformation, for 36–48 h (Nabi et al. 1993).

Primary RPE cultures were prepared from 2-wk-old Long Evans rats (Harlan Sprague Dawley). Animals were killed by CO₂ asphyxiation, the eyes were enucleated and stored in HBSS. A circumferential incision was made above the ora serrata, and the cornea, lens, iris, and vitreous body were removed. The eyecups with the neural retina exposed were incubated in 320 U/ml hyaluronidase in HBSS for 1 h at 37°C, the neural retina was peeled off from the RPE, and the eyecups were incubated in 2 mg/ml trypsin in HBSS for 60 min at 37°C. RPE sheets were teased from the underlying choroid with needles, collected, and incubated with trypsin/EDTA for 1 min. The cells were plated on Matrigel-coated Transwell filters and cultured without further passaging in DME supplemented with 10% FCS, l-glutamine, nonessential amino acids, and antibiotics.

Phosphorothioate antisense oligonucleotides used in this study were complementary to the position −4 to 20 (relative to the translation initiation site) of rat ezrin mRNA (GenBank/EMBL/DDBJ, X67788; Barila et al. 1995). Oligonucleotides were purchased from Cornell University and Operon Technologies, Inc. Sense oligonucleotides corresponding to positions −4 to 20 were used as controls. The antisense oligonucleotide sequence was analyzed with the BLAST sequence database and did not show significant homology with other known sequences. Primary RPE cells on 6.5-mm diam Transwell filters were treated with oligonucleotide as described by Takeuchi et al. 1994. In brief, 20 μM oligonucleotides were added into the culture medium of the apical chamber every 4 h for 96 h. Every four additions, the culture medium was exchanged with fresh medium containing 20 μM oligonucleotide. The antisense effects described were neutralized by the addition of the same concentration of sense oligonucleotide.

VSVG-tagged cDNA coding for the wild-type human ezrin, a generous gift of Dr. Monique Arpin (Curie Institute, Paris, France), was previously described (Algrain et al. 1993). RPE-J cells were seeded on 10-cm petri dishes 24 h before transfection using a standard calcium precipitate protocol (Chen and Okayama 1988). Transfected cells were selected in medium containing 0.6 mg/ml G-418. Three clones, which produced the exogenous ezrin at different levels (as determined by immunoblots) were selected for further studies.

CO₂ anesthetized Long Evans rats of different ages had an intracardiac perfusion with 4% paraformaldehyde made in PBS supplemented with 0.3 mM CaCl₂ and 1 mM MgCl₂ (PBS/CM). Their eyes were enucleated and postfixed by immersion in the same fixative overnight. Eyecups were infused successively with 10 and 20% sucrose in PBS/CM, and with Tissue-Tek “4583” (Miles Inc.). 10-μm cryosections were cut on a cryostat (Bright Instrument Company) and collected on SuperFrost slides (Fisher).

Cells on filters were fixed in 4% paraformaldehyde in PBS/CM for 20 min, quenched in 50 mM NH₄Cl in PBS/CM for 20 min, and permeabilized in 0.2% Triton X-100 in PBS/CM for 10 min. After blocking in PBS/CM supplemented with 0.2% BSA (PBS/CM/BSA) for 30 min, the filters and cryosections were incubated with the ezrin monoclonal or polyclonal antibodies, moesin pAb, or VSVG mAb in PBS/CM/BSA for 1 h. The samples were washed in PBS/CM/BSA and incubated with secondary antibodies coupled to FITC or Cy™3 for 45 min. Cell nuclei were labeled with 1 μg/ml 4′,6-diamidino-2-phenylindole in PBS/CM for 5 min, or with 2 μg/ml propidium iodide in PBS/CM for 15 min. 1-μm X-Y (en face) or X-Z (transverse) sections were obtained using a dual channel laser scanning confocal microscope (Sarastro, Molecular Dynamics). Alternatively, samples were analyzed using an epifluorescence microscope (model E600; Nikon). Digital images were collected with a cooled CCD camera and Metamorph software (Universal Imaging). Microscopic panels were composed using Adobe Photoshop 5.0.

Freshly isolated rat eyecups of different ages, primary RPE, RPE-J, and RPE-J clones plated on filters were fixed in 2.5% glutaraldehyde, 0.2% picric acid in 0.1 M cacodylate buffer, pH 7.3. Samples were postfixed in 1% osmium tetroxide for 1 h, dehydrated in acetone and embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate. Specimens were examined at 80 kV in a JEOL 100 CXII electron microscope. For scanning EM, samples were plated on 1% Alcian blue-treated (Sigma Chemical Co.) glass coverslips, fixed in 2.5% glutaraldehyde freshly diluted in 0.1 M cacodylate buffer, postfixed in 1% osmium tetroxide for 30 min, dehydrated in graded ethanol solutions, critical point-dried in liquid carbon dioxide, coated with vacuum-evaporated gold-platinum, and examined in a JEOL 100 CXII microscope equipped with the ASID high resolution scanning module, using an accelerating voltage of 20 kV.

Freshly isolated rat eyecups of adult rats were fixed in 4% paraformaldehyde, 0.1% glutaraldehyde, 0.2% picric acid in 0.1 M modified PHEM buffer, pH 6.9 (60 mM Pipes, 20 mM Hepes, 10 mM EGTA, 5 mM MgCl₂, 70 mM KCl), for 1 h at 4°C. Samples were postfixed in 0.25% tannic acid for 1 h at 4°C, dehydrated in ethanol, embedded in Unicryl Kit resin (Ted Pella Inc.), and polymerized under UV light for 96 h at −20°C. Ultrathin sections on grids were hydrated at room temperature for 1 h with PBS, 1% BSA, 0.01% Tween 20, 10% goat serum, and incubated in 50 mM NH₄Cl in PBS/CM for 30 min, followed by sequential incubation with antiezrin rabbit antibody (1:100) and 10-nm colloidal gold-conjugated antibody (Ted Pella Inc.) for 1 h each.

Whole cell lysates from RPE-J monolayers, RPE collected from rat eyecups, and primary cultures of RPE were solubilized in PBS supplemented with 2% SDS and a cocktail of protease (1 mM PMSF, 20 μg/ml leupeptin, 20 μg/ml pepstatin A, and 40 μg/ml aprotinin) and phosphatase (100 μM sodium vanadate) inhibitors. 10 μg of total protein of each sample was resolved in a 7.5% SDS-PAGE and electrotransferred to nitrocellulose membranes (Schleicher & Schuell). To detect ezrin and moesin, membranes were incubated with polyclonal ezrin antibody C-19 in Blotto A buffer (20 mM Tris/HCl, 0.9% NaCl, 0.05% Tween 20, 5% skimmed milk) as directed by the manufacturer. Protein detection was performed with secondary antibodies conjugated to peroxidase and visualized using the Western blot chemiluminescence Reagent Plus (New England Nuclear Life Science Products, Inc.) detection system. The intensity of scanned protein bands was analyzed with NIH Image 1.62 software.

Association of ezrin with the cytoskeleton was defined by immunofluorescence and cell fractionation into detergent soluble and insoluble fractions as previously described (Crepaldi et al. 1997; Kreis 1987). For immunofluorescence, plated cells were washed in PBS/CM and dipped four times for 5 s at room temperature in four different beakers containing extraction buffer MES (50 mM MES, 3 mM EGTA, 5 mM MgCl₂, 0.5% Triton X-100, pH 6.4), fixed in 4% paraformaldehyde, and processed for immunofluorescence as described above. Alternatively, cells seeded on 6.5-mm Transwell filters were extracted in 300 μl extraction buffer MES for 40 s at room temperature. Detergent-soluble fractions were precipitated in 80% acetone for at least 4 h at −20°C, and centrifuged at 300 g for 10 min at 4°C. Detergent soluble and insoluble fractions were resuspended in 30 μl of Laemmli sample buffer each, loaded, and analyzed by 7.5% SDS-PAGE and immunoblotting.

RPE cells and the neighboring PR undergo dramatic shape changes during postnatal maturation of the retina in the rat. Histological and fluorescence analysis of the immature rat eyecup at postnatal day 2 (P2; Fig. 1, A–C) shows RPE cells in direct contact with the immature PR, identified by their nuclei (stained blue in Fig. 1 A, and red [with propidium iodide] in Fig. 1B and Fig. C), as the outer segments of the PR have not developed yet (see Fig. 1, A–C). In contrast, in animals two weeks of age or older, the mature retinal architecture and interactions have been established (Fig. 1, D–F) and the RPE is separated from the outer nuclear layer and the inner PR segments by well developed PR outer segments. Immunofluorescence of rat eyecup cryosections detected ezrin at the apical surface of both immature (Fig. 1 B) and adult (Fig. 1 E) RPE cells. Upon RPE maturation, apical ezrin staining was more intense; in the adult retina, it was observed in the external regions of the PR layer (Fig. 1 E), where RPE microvilli ensheath the fully developed outer segments. Parallel cryosections of P2 (Fig. 1 C) and adult (Fig. 1 F) retina stained with a specific moesin antibody failed to detect it at the RPE layer. Some moesin staining was detected in the capillaries at the choroid layer underneath the RPE layer at both stages of development. Moreover, the labeling of the same cryosection with pAb 457 specific to mouse radixin failed to detect it at the RPE layer (data not shown).

The structural changes associated with RPE maturation, visible by EM, are shown in Fig. 2. The RPE of the newborn rat displays short and stubby apical microvilli and very short invaginations and tubules in association with its basal plasma membrane (Fig. 2 A). By P7, half-way through maturation, the apical surface has extended long microvilli whereas the basal surface has developed immature basal infoldings (Fig. 2 B). By P14, rat RPE displays apical microvilli and maximally developed basal infoldings (data not shown), identical to those found in the RPE of the adult rat (Fig. 2 C). Previous morphometric studies of rat eyecups of different ages have shown that both apical and basolateral surfaces of RPE double their area during postnatal maturation (Marmorstein et al. 1998).

Postembedding immunocytochemistry of adult eyecups was used to further examine the ultrastructural localization of ezrin in RPE cells. Ultrathin sections of adult eyecups embedded in unicryl were sequentially labeled with an affinity-purified pAb to ezrin, followed by a secondary donkey anti–rabbit IgG antibody conjugated with colloidal-gold particles. In these samples, labeling was specifically associated with actin bundles in apical microvilli (Fig. 3 A, arrowheads) and, to a lower extent, with basal infoldings (Fig. 3 B). Importantly, no ezrin labeling was associated with the apical plasma membrane adjacent to the microvilli. A similar ezrin distribution was observed in primary cultures of mature RPE cells plated on semipermeable polycarbonate filters (data not shown). Control samples labeled with secondary antibody alone showed no labeling (data not shown). These data demonstrate that ezrin has a dual apical and basal distribution in RPE cells in vivo, as in parietal cells of gastric glands (Hanzel et al. 1991), but, unlike most other epithelia, where it is usually only found at apical microvilli (Berryman et al. 1993).

If ezrin is involved in the process of RPE maturation, it may be upregulated during this time. Immunoblots of lysates of RPE cells obtained from rats of different postnatal ages demonstrated a progressive increase in ezrin with adult levels observed after two weeks (Fig. 4 A). Quantitation of these blots showed that ezrin expression was upregulated ∼400% between P2 and adults (Fig. 4 B). In contrast, moesin and radixin, also reactive with the pAb used in these studies, were not detected in either developing or adult RPE. These data are in agreement with our immunofluorescence data which detected neither moesin (Fig. 1) nor radixin (data not shown) in RPE in vivo. The large increase of ezrin expression during postnatal maturation of RPE further suggested that this protein may indeed play an important role in the establishment of apical and basal actin-based structures.

Primary RPE cultures preserve long microvilli and basal infoldings characteristic of adult RPE in the adult eye (Stramm et al. 1983; Heth et al. 1987; Gundersen et al. 1993; Davis et al. 1995). Even when RPE in situ expressed only ezrin, we tested the primary RPE cultures for the expression of other ERM proteins, as it is well known that their expression is more promiscuous under culture conditions. Primary RPE monolayers were probed with an antibody specific to a COOH-terminal peptide conserved in all three ERM proteins. Whole extracts of these cells revealed the presence of two bands corresponding to the molecular weight of ezrin and moesin (Fig. 5 A, lane 1, arrows); no band with the molecular weight of radixin (intermediate between ezrin and radixin) was detected. Quantification of ezrin and moesin from the Western blot indicated a 1.7:1 ratio between these two proteins. To test the association of ezrin and moesin with the cytoskeleton, we carried out detergent extractions. 45 ± 1.0% of ezrin and 52 ± 3.5% of moesin in primary RPE were extracted (Fig. 5 A, lanes 2 [soluble] and 3 [insoluble]), indicating that a large fraction of each protein was apparently linked to the cytoskeleton.

We analyzed the distribution and detergent extractability of ezrin and moesin in primary RPE monolayers by immunofluorescence (Fig. 5, B–I). Confocal microscopy en face examination of paraformaldehyde-fixed monolayers grown on polycarbonate filters revealed an apical punctate staining for ezrin consistent with an apical localization, similar to that observed in RPE in situ (Fig. 5 B). Cross-sections of these monolayers confirmed the apical localization of ezrin (Fig. 5 C). Confocal immunofluorescence analysis of primary RPE cultures first extracted with 0.5% Triton X-100 and then fixed with paraformaldehyde, revealed the presence of ezrin, primarily in apical regions consistent with the association with apical microvilli (Fig. 5D and Fig. E). On the other hand, moesin appeared to be associated with apical, lateral, and basal surfaces of primary RPE cells fixed before extraction (Fig. 5F and Fig. G), but only on the basal surface of cells extracted with 0.5% Triton X-100 before fixation with paraformaldehyde (Fig. 5H and Fig. I). Incubation of primary RPE monolayers with an antibody specific to radixin failed to stain the cells (data not shown). The data suggest that in primary RPE cultures, a major pool of ezrin is associated with the apical surface of the cells, while the most stable pool of moesin interacts with basolateral membranes.

To test the hypothesis that ezrin is an important player in the establishment of apical microvilli and basal infoldings, we used an antisense oligonucleotide approach. Primary RPE cultures were exposed for 96 h to antisense oligonucleotides directed to the first 24 amino acids encoded by the rat ezrin mRNA (Fig. 6). In control RPE cultures not exposed to antisense oligonucleotides, the expression level of ezrin, measured by immunoblot, was high, similar to that of RPE cells obtained directly from the retina for analysis (Fig. 6 A, lane 1). As previously shown, moesin was also expressed. Importantly, addition of sense oligonucleotides did not affect either ezrin or moesin expression in these cells (Fig. 6 A, lane 2). In contrast, primary RPE monolayers cultured in the presence of ezrin antisense oligonucleotides displayed an almost complete inhibition of ezrin expression relative to untreated control cells, whereas moesin expression was not affected or was inhibited to a much smaller extent (Fig. 6 A, lane 3). In control experiments, addition of sense plus antisense oligonucleotides directed to ezrin resulted in levels of expression of both ezrin and moesin similar to control levels (data not shown). The effects of ezrin antisense oligonucleotides were also examined by immunofluorescence (Fig. 6, B–G). Control primary RPE cells displayed a typical microvillar apical staining pattern for ezrin, consisting of punctate or short linear elements frequently coalescing into clumps (Fig. 6 B). Instead, moesin expression was diffuse throughout the cytoplasm with small puncta of undefined origin not associated with the apical surface (Fig. 6 C). No change in the ezrin (Fig. 6 D) or moesin (Fig. 6 E) labeling patterns was observed when monolayers were supplemented with ezrin sense oligonucleotides or with a mixture of sense and antisense oligonucleotides (data not shown). In sharp contrast, monolayers supplemented with ezrin antisense oligonucleotides almost completely lost the ezrin staining (Fig. 6 F), whereas moesin staining remained largely unchanged (Fig. 6 G).

The immunofluorescence observations were confirmed by both scanning (Fig. 7A, Fig. C, and Fig. E) and transmission (Fig. 7B, Fig. D, and Fig. F) EM. Untreated primary RPE cultures displayed an apical surface densely covered by microvilli (Fig. 7A and Fig. B) and a basal surface with elaborate basal infoldings (Fig. 7 B). Control monolayers treated with ezrin sense oligonucleotides were essentially unchanged (Fig. 7C and Fig. D). Strikingly, treatment of primary RPE cultures with ezrin antisense oligonucleotides caused almost complete disappearance of microvilli (Fig. 7E and Fig. F) and of basal infoldings (Fig. 7 F). Observation at higher magnification (Fig. 7, insets) demonstrated that microvilli in monolayers treated with antisense oligonucleotides (Fig. 7 E) were very short and sparse (Fig. 7 E), in stark contrast with the long and abundant microvilli of control (Fig. 7 A) and sense oligonucleotide-treated (Fig. 7 C) samples. The data are consistent with a stringent requirement for high levels of ezrin protein and with a limited role of moesin in the maintenance of apical RPE microvilli and basal infoldings.

To further analyze the morphogenetic role of ezrin in RPE cells we analyzed ezrin distribution and levels of expression in RPE-J cells, a rat cell line established in our laboratory that preserves many native RPE characteristics (Nabi et al. 1993; Bonilha et al. 1997), including the ability to perform phagocytosis of outer segments (Finnemann et al. 1997). In contrast with primary RPE cultures, RPE-J cells possess sparse and short microvilli at their apical surface and no basal infoldings (Nabi et al. 1993; Marmorstein et al. 1998). Confocal microscopy examination of paraformaldehyde-fixed RPE-J monolayers grown on polycarbonate filters revealed a diffuse cytoplasmic distribution of ezrin (Fig. 8A and Fig. B). Moesin staining appeared to be slightly more granular and also extended throughout the cytoplasm (Fig. 8 E). Examination of cross-sections of these monolayers showed that moesin was evenly distributed over the apical and basolateral surfaces of RPE-J cells (Fig. 8 F). To test whether the diffuse cytoplasmic distribution of ezrin in RPE-J cells reflected a weak or absent interaction of ezrin with the actin cytoskeleton, we carried out standard detergent extractability assays. Extraction of monolayers with 0.5% Triton X-100 and fixation with paraformaldehyde removed most of the ezrin in RPE-J cells (Fig. 8C and Fig. D), suggesting that in these cells, ezrin is not anchored to the actin cytoskeleton. In contrast, a larger fraction of moesin staining was preserved in monolayers treated with detergent and appeared to be concentrated at the base of the cells, below the nuclei (Fig. 8G and Fig. H).

To provide additional evidence for the role of ezrin in the assembly of RPE microvilli and basal infoldings, we generated stably transfected RPE-J cells overexpressing human ezrin tagged with a VSVG cytoplasmic epitope (Algrain et al. 1993). Immunoblot experiments indicated that RPE-J cells express no radixin and low levels of ezrin and moesin (ratio 1.15:1), comparable to immature (P2) RPE (Fig. 4) and three to four times lower than primary cultures of adult RPE cells (Fig. 9 A). RPE-J clones 6, 16, and 20 overexpressed VSVG-tagged ezrin at levels two to four times higher than wild-type RPE-J (Fig. 9 A). Interestingly, these RPE-J clones displayed a significantly higher proportion of cytoskeleton-associated ezrin than wild-type RPE-J cells (63 ± 8.6%, 58 ± 3.5%, and 69 ± 6.8%, for clones 6, 16, and 20, compared with 37 ± 2.1% in wild-type RPE-J, as determined by resistance to Triton X-100 extraction; Fig. 9 B). Immunofluorescence experiments of monolayers sequentially extracted and fixed showed that >70% of the cells in these RPE-J clones expressed transfected ezrin and displayed an apical microvilli pattern resistant to detergent extraction (Fig. 9E, Fig. G, and Fig. I) that was not observed in control RPE-J cells (Fig. 8, A–D). Importantly, moesin levels were not altered (Fig. 9 A) and its partition into the detergent-resistant fraction was not significantly increased (Fig. 9 B) by ezrin overexpression (53 ± 6.9%, 47 ± 5.6%, and 40 ± 7.3% in clones 6, 16, and 20, respectively, compared with 57 ± 3.8% in control RPE-J cells). Overexpression of ezrin appeared to promote a subtle change in moesin distribution, from a preferentially basolateral distribution (Fig. 9 D) to a more diffuse cytoplasmic staining (Fig. 9F, Fig. H, and Fig. J).

The ultrastructure of RPE-J clones overexpressing the human ezrin cDNA was analyzed by scanning (Fig. 10A, Fig. C, and Fig. E) and transmission (Fig. 10B, Fig. D, and Fig. F) EM. As previously shown (Nabi et al. 1993), RPE-J cells display very short and scattered microvilli at their apical surface (Fig. 10A and Fig. B) and no basal infoldings (Fig. 10 B). Fig. 10 shows electronmicrographs of clones 6 (Fig. 10C and Fig. D) and 20 (Fig. 10E and Fig. F). RPE-J cells transfected with ezrin exhibited an increase in the number and length of their apical microvilli (Fig. 10C and Fig. E, see corresponding insets). Fig. 10 D illustrates the presence of basal infoldings in clone 6. Taken together, the data presented in this section demonstrate that overexpression of ezrin in RPE-J cells is sufficient to induce both apical microvilli and basal infoldings.

The results of the experiments in this report provide strong evidence for a specific morphogenetic role of ezrin in RPE. Moesin appears not to be required for these morphogenetic events, as it is not expressed by RPE in vivo, and is largely not affected by the two in vitro strategies employed. The third member of the ERM family, radixin, was not detected in either RPE in situ or under culture conditions, indicating that, as is the case with moesin, it is not normally involved in these processes.

The first important evidence in support of a specific morphogenetic role of ezrin was provided by immunolocalization studies in rat eyes. Colloidal gold immunoelectron microscopy decorated ezrin along the entire actin filamentous core of RPE microvilli and throughout the convoluted basal infoldings (Fig. 3). The data extend prior observations that demonstrated the presence of ezrin in RPE microvilli (Hofer and Drenckhahn 1993). As in most other epithelia (Berryman et al. 1993), ezrin was accompanied by no other ERM protein. Only a few epithelial tissues display ezrin and moesin in microvilli and, in just one case (the hepatic bile canaliculi), moesin is the only ERM protein expressed in epithelial microvilli (Berryman et al. 1993). On the other hand, moesin, but not ezrin, is expressed in endothelial cells. A basolateral localization of ezrin has previously been reported for immature enterocytes in intestinal crypts, for the intricate podocyte extensions to the glomerular basement membrane in renal corpuscles (Berryman et al. 1993), and for the basolateral membrane infoldings of both resting and stimulated parietal cells (Hanzel et al. 1991). The localization of ezrin at apical and basolateral membranes of RPE cells is likely to depend on the interaction with different plasma membrane proteins. In fact, studies carried out in parietal cells showed that ezrin can bind to proteins that are localized either apically or basolaterally (Hanzel et al. 1991; Yao et al. 1993, Yao et al. 1996). Experiments in other cell systems have identified several plasma membrane proteins that bind the NH₂-terminal domain of ezrin, including CD44, CD43, proton transporting ATPase, ICAM-1, ICAM-2, and peripheral membrane proteins (such as EBP50) that may act as a bridge between ezrin and the plasma membrane. It is not known at present which RPE proteins anchor ezrin to the apical or to the basal plasma membrane of RPE.

A second important piece of evidence for a role of ezrin in RPE morphogenesis was provided by developmental studies that showed a direct correlation between spacial and temporal increase in ezrin expression and the development of surface differentiations of RPE. Immunoblots of whole RPE cell lysates at different days after birth demonstrated that the ezrin levels increased fourfold as the RPE developed microvilli and basal infoldings and established adult interactions with PRs and the basement membrane. This correlates well with the reported fourfold increase in actin content during maturation of chicken RPE (in chicken, as in humans, the maturation of the retina occurs before birth; Owaribe and Eguchi 1985), which reflects the structural need for these two proteins in the assembly of microvilli and basal infoldings. As immunoblots of RPE lysates obtained from rats of different ages did not detect the presence of moesin or radixin (Fig. 4), we can discard a possible additional role of these proteins in vivo. Ezrin has been identified as a minor component of apical brush borders of intestinal and renal epithelial cells and a previous report described an increase in ezrin mRNA as intestinal cells matured from an embryonic stage to a mature morphologically differentiated stage (Barila et al. 1995). However, in intestinal and kidney epithelia, potential morphogenetic roles of ezrin and other ERM proteins are overshadowed by the contribution of villin, a major component of microvilli that is a major player in the assembly of these structures.

A third piece of evidence in support of a morphogenetic role of ezrin was provided by antisense oligonucleotide experiments. We speculated that, since villin is not expressed by RPE (Hofer and Drenckhahn 1993), the possible contributions of ezrin might be relatively more important and easier to detect in this model system. Experiments in which ezrin antisense oligonucleotides were added to primary cultures of mature RPE cells, which preserve apical microvilli and basal infoldings (Stramm et al. 1983; Heth et al. 1987; Gundersen et al. 1993; Davis et al. 1995), yielded a striking result: 80% drop in ezrin levels resulted in a striking decrease in the number and length of RPE microvilli and the complete disappearance of basal infoldings. Our studies markedly differ from two previous reports in which an antisense approach was used to explore the function of ERM proteins. In the experiments by Takeuchi et al. 1994 and Paglini et al. 1998, a mixture of antisense nucleotides against two to three ERM proteins was required to observe a clear morphogenetic effect on, respectively, microvilli or growth cone extension; in our experiments with RPE cells, antisense oligonucleotides against just ezrin cause a dramatic morphological effect. Ezrin antisense oligonucleotides did not affect, or decreased only marginally, moesin expression. Furthermore, since moesin was preferentially associated with the basolateral membrane, the data does not support an important role of this protein in the development and maintenance of RPE microvilli. In addition, since moesin was not detected in RPE in vivo, it is obviously not involved in the morphogenesis of basolateral infoldings in vivo, although it might contribute to their formation in vitro.

A fourth important piece of evidence for a role of ezrin in RPE morphogenesis was provided by transfection and overexpression of ezrin cDNA into cultures of the rat RPE cell line, RPE-J. Although this cell line preserves many native RPE functions, such as similar targeting and sorting of some plasma membrane proteins and the phagocytosis of PR outer segments (Nabi et al. 1993; Bonilha et al. 1997; Finnemann et al. 1997; Marmorstein et al. 1998), it does not preserve the characteristic apical microvilli and basal infoldings of RPE in situ. In contrast with primary RPE cultures, which displayed high levels of insoluble ezrin preferentially associated with the apical microvilli, RPE-J cells had three- to fourfold lower ezrin levels, a diffuse cytoplasmic distribution, and a larger proportion of detergent-soluble ezrin, indicative of a poor association with the actin cytoskeleton. As in primary RPE cultures, moesin was preferentially associated to the basolateral cytoskeleton of RPE-J cells. Because the data suggested that higher levels of expression of ezrin associated with the cytoskeleton is an important factor in the establishment of apical microvilli and basal infoldings in RPE, we studied the effect of stable ezrin overexpression by transfection into RPE-J cells. Analysis of clones with different levels of ezrin expression detected a clear correlation between high levels of ezrin expression and the formation of apical microvilli and basal infoldings. Furthermore, overexpressed ezrin showed an increased resistance to detergent extraction and microvillar localization.

Our transfection data contrast with previous studies in which ezrin overexpression failed to observe any striking modification of cell surface morphology. Transfection of full-length ezrin cDNA into monkey kidney cells resulted in the accumulation of ezrin underneath the plasma membrane in actin-containing membrane protrusions, such as microspikes, filopodia, and microvilli, with no obvious reorganization of actin-containing structures (Algrain et al. 1993). Microinjection of ezrin protein into living gastric cells resulted in the accumulation of the protein at microvilli and other membrane locations without inducing morphological changes (Andreoli et al. 1994). On the other hand, overexpression of the COOH-terminal domains of ezrin and radixin, which bind to F-actin, induced clear surface alterations (Henry et al. 1995; Martin et al. 1995, Martin et al. 1997), whereas overexpression of the NH₂-terminal domain of ezrin and moesin reduced the size and number of microvilli in LLC-PK cells (Algrain et al. 1993; Crepaldi et al. 1997) and interfered with surface structures of fibroblasts (Amieva et al. 1998). The lack of effect of transfected full size ezrin in some cell systems may reflect its inactivated monomer state, and the absence of cellular activating mechanisms. Our transfection experiments suggest that RPE-J cells retain some capacity to activate ezrin into the assembly of microvilli. However, as the induced microvilli are relatively short compared with those present in native adult RPE, additional factors needed for the assembly of RPE microvilli may be in short supply in RPE-J cells or the balance of ezrin activation and deactivation systems may not be the same as in native RPE. As RPE-J cells express lower ezrin levels than primary RPE cultures and many other cell lines, ezrin expression in RPE-J cells may be below the threshold necessary to ensure cell spreading and morphological changes driven by exchanges between the soluble and membrane–cytoskeleton-associated pools.

The results of this report suggest that moesin expression cannot substitute for ezrin in the generation of apical microvilli and basal infoldings in RPE cells, thus supporting a very specific role of ezrin in these processes. Our results suggest that RPE cells express specific ezrin binding proteins in the apical surface which cannot bind moesin as efficiently. Moesin, on the other hand, may bind other plasma membrane proteins which are not essential in the establishment or maintenance of RPE microvilli or basal infoldings. It is quite clear that ERM proteins bind specific motifs in plasma membrane proteins and that this binding is essential for their morphogenetic effects (Yonemura et al. 1999). The apparent specificity of ERM protein function supported by our data apparently conflicts with the observation that moesin knockout mice fail to show obvious histological abnormalities, or deficiencies in various cellular processes previously attributed to moesin, which implies functional redundancy of ERM proteins at both cellular and whole body level (Doi et al. 1999). Since this study excluded the eye, future experiments in which ezrin expression in the eye is blocked in vivo should help provide an answer to this paradox.

The dramatic changes in ezrin expression during developmental maturation and the striking effects of ezrin antisense and overexpression turn RPE into an excellent model to study in detail the function of ezrin and the associated proteins in vivo and in vitro. The uniqueness of the RPE system may lie in the special complement of cytoskeletal proteins expressed by these cells. Our results with both the suppression of ezrin expression in primary cultures and the overexpression of ezrin in RPE-J cells are very similar to previous reports about villin function in the generation of brush border microvilli (Friederich et al. 1989; Costa de Beauregard et al. 1995). However, whereas villin participates in the Ca²⁺-dependent bundling of actin filaments, there is no evidence that ezrin could perform a similar role, even in a Ca²⁺-independent manner. In this regard, ezrin interaction with actin filaments may be much more complex than originally imagined. Recent work has indicated that ezrin has the ability to stimulate actin polymerization and preferentially bind to nonmuscle β-actin isoforms in vitro (Yao et al. 1996). Furthermore, an actin binding site which mediates binding to both G-actin, as well as to F-actin, has been mapped to ezrin residues 281–310 (Roy et al. 1997) and the sequence between amino acids 558–578 of ezrin, radixin, and moesin is highly homologous to the actin binding site of the barbed-end capping protein CapZ β subunit (Turunen et al. 1994). Taken together, these data suggest that ezrin and the other ERM proteins may also have barbed-end capping activity in vivo, thus regulating the length of actin filaments by both stimulating and inhibiting actin polymerization. Ongoing work in our laboratory aims to elucidate the molecular components and signaling pathways that regulate ezrin function in RPE cells.

The authors are grateful to Dr. Monique Arpin (Institut Curie, Paris, France) for providing us with the pAb against ezrin and the human VSVG-tagged ezrin cDNA, and to Drs. Anthony Bretscher (Cornell University, Ithaca, NY) and Frank Solomon (Massachusetts Institute of Technology, Cambridge, MA) for their kind gifts of antibodies against moesin and radixin. We thank Leona Cohen-Gould and Dena Almeida for excellent assistance with the EM and cryosectioning of rat eyes.

This work was supported by the National Institutes of Health grant EY08538, and a Jules and Doris Stein professorship from the Research to Prevent Blindness Foundation to E. Rodriguez-Boulan, and a Norman and Rosita Winston Fellowship in Biomedical Research to S.C. Finnemann.